All datasets needed to run our notebooks can be found here. We used several additional packages in the sentiment analysis section. We created a shell script that downloads and updates and required packages and sets up the datasets folder. The script can be found here and running the following command should set everything up and start a Jupyter notebook session on the CS109b AMI with a token instead of a password:
wget https://s3.amazonaws.com/109b/get_datasets.sh && . get_datasets.sh
All milestone notebooks were submitted over the course of the project. The remaining notebooks were created for the final submission.
- Milestone 1: initial data exploration
- Milestone 2: assembling training data
- Milestone 3: traditional methods
- Milestone 4: deep learning models
- Additional data exploration: metadata pairplots and sentiment boxplots
- Additional sentiment analysis: most frequent words in the TMDb dataset
- Genre frequencies: comparison between original and predicted genre frequencies
- Keras with metadata: deep learning models with movie metadata instead of posters
- Predictions heatmap: heatmap of predicted genres